AI Agent: Plan Prom Without the Stress, Part 2

Prom has a thousand moving parts: outfit, group, dinner res, transport, afterparty. An AI agent can hold the whole plan in its head and update it when one piece changes.

Some examples

• Agent suggests dinner spots near your venue with group seating.
• Agent tracks who in the group still hasn't paid for the limo split.
• Agent drafts a group chat update when plans shift.
• Agent keeps a budget so the whole night doesn't blow your savings.

Try it!

Have an AI agent plan a fake prom for 8 friends, $150 each, with dinner and afterparty. Check the math.

SAT prep works when it's daily and targeted. An AI agent picks today's drill, reviews yesterday's misses, and tracks which question types still trip you up.

Some examples

• Agent sends a 20-minute daily set tuned to your weakest section.
• Agent reviews your wrong answers and explains the rule.
• Agent re-quizzes you on missed concepts a week later.
• Agent shows a streak chart so you don't break the chain.

Try it!

Set up an AI agent to give you 5 SAT questions a day for one week. Track your improvement.

Reselling clothes can pay for your phone bill. An AI agent helps you spot what to buy, write listings that convert, and price each piece based on real comps.

Some examples

• Agent estimates resale price from a thrift-store photo.
• Agent drafts a Depop listing with hashtags and measurements.
• Agent tracks your inventory and total profit per piece.
• Agent flags slow-moving items and suggests a price drop.

Try it!

Take a photo of one item in your closet. Have an AI agent draft a real listing and pick a price.

Lab reports follow a strict format: hypothesis, methods, results, discussion. An AI agent reads your rubric, asks for your raw data, and helps you draft each section without doing it for you.

Some examples

• Agent quizzes you on what your hypothesis actually predicted.
• Agent helps you turn a messy data table into a clean graph caption.
• Agent flags when your discussion makes claims your data can't support.
• Agent checks your draft against the actual rubric you upload.

Try it!

Give an AI agent a real rubric and your raw data. Let it coach you through one section.

Launching means a hundred small tasks: post on socials, message your list, fix the one bug you missed. An AI agent runs the checklist, drafts the posts, and pings you when something stalls.

Some examples

• Agent drafts a launch post for X, LinkedIn, and Reddit in different tones.
• Agent writes a Show HN draft and warns you about Hacker News etiquette.
• Agent monitors form submissions and alerts you on the first signup.
• Agent reminds you to thank early users by name.

Try it!

Pick a project (real or fake). Have an AI agent build a 12-step launch checklist with draft posts.

Designing a game alone is hard — you can't really playtest your own ideas. An AI agent can roleplay as a player, find broken combos, and suggest tweaks before you ever write code.

Some examples

• Agent simulates 20 fights and reports the win rates of each character.
• Agent finds the dominant strategy nobody intended.
• Agent generates 10 enemy variants from one base type.
• Agent writes the tutorial level script in your tone.

Try it!

Describe a tiny dice game in 3 sentences. Have an AI agent simulate 50 plays and report the imbalance.

Yearbook means chasing 200 people for quotes, photos, and approvals. An AI agent tracks who's done, drafts caption variations, and reminds you about deadlines without being annoying.

Some examples

• Agent generates 5 caption options for a candid photo.
• Agent tracks which seniors haven't submitted quotes.
• Agent drafts personalized reminder messages to chase missing photos.
• Agent checks your spread for inclusivity gaps before print.

Try it!

Give an AI agent a fake list of 10 students missing quotes. Have it draft 10 personal nudges.

Being club treasurer means receipts, spreadsheets, and reimbursements. An AI agent reads your statements, categorizes expenses, and drafts the report your advisor needs at year end.

Some examples

• Agent extracts amounts and dates from photographed receipts.
• Agent flags when you're trending over budget mid-semester.
• Agent drafts the annual report with charts.
• Agent reminds members to submit reimbursement forms.

Try it!

Snap photos of 5 fake receipts. Have an AI agent extract amounts, categorize, and total them.

Finding an internship as a teen is a search problem. An AI agent crawls listings, filters by 'open to high schoolers', tracks deadlines, and tailors a cover letter to each one.

Some examples

• Agent surfaces 10 internships matching your interests this week.
• Agent tailors your cover letter to each company's actual mission.
• Agent tracks status: applied, interviewed, ghosted, offered.
• Agent reminds you to follow up after 7 days of silence.

Try it!

Pick a field. Have an AI agent find 5 real internships open to teens and draft tailored cover letters.

ChatGPT in a chat window is a chatbot — it talks. ChatGPT with browsing, file access, and the ability to run code is an agent — it acts. Agents loop: they pick a tool, use it, look at the result, and pick again until the task is done.

Some examples

• A chatbot tells you the weather; an agent books the umbrella delivery.
• Claude in chat answers questions; Claude Code edits your files and runs your tests.
• ChatGPT explains how to send an email; an agent with email tools actually sends it.
• A chatbot summarizes a webpage you paste; an agent fetches the page itself, then summarizes.

Try it!

Open ChatGPT or Claude. Ask the same task two ways: once requiring it to act (use a tool), once just to explain. Notice which feels more useful for what you want.

An agent isn't magic — it's an LLM in a loop with tools. Give Claude a goal like 'organize my tasks by deadline' and tools to read and write your to-do file, and it'll plan, act, and check itself without you babysitting every step.

Some examples

• You give Claude readFile and writeFile tools plus 'sort tasks.md by due date' and it does it in three loop iterations.
• An agent with a search tool and a goal of 'find unfinished TODOs in src/' returns a single grouped report.
• Claude with a calendar tool plans 'block focus time tomorrow morning' across three back-and-forth turns.
• An agent with shell access asked to 'find duplicate images' chooses to use sha256sum instead of pixel comparison.

Try it!

Give Claude two tools (read and write a file) and a goal — 'reorganize my notes by topic'. Watch it plan and act.

An AI agent doesn't just respond to you — it takes actions in the real world: sending emails, booking appointments, executing code, controlling browsers. The difference matters because the consequences of mistakes get bigger. Knowing what agents are now and what they can already do helps you use them well and stay safe.

Some examples

• Browser-using agents can fill out forms and click buttons on real websites.
• Email agents can draft, send, and reply on your behalf.
• Coding agents can write, test, and deploy code without your input on each step.
• Operator agents from major labs can do multi-step real-world tasks.

Try it!

Read about one agent product (Claude Computer Use, OpenAI Operator). Note what it can and can't do.